Research On Tongue Ecchymosis Recognition And Tongue Image Acquisition Device Improvement

Traditional Chinese medicine(TCM)is a gem of Chinese traditional culture.Its "unity of man and nature" and the syndrome differentiation system make people pay more and more attention to it.However,the inheritance and development of tongue diagnosis,which is an important part of TCM,is restricted by its subjectivity.Therefore,the standardization,quantitation and objectivityof tongue diagnosis must be the inevitable trend of future development.The subject studied the objectification of tongue ecchymosis.The main research contents are tongue image segmentation,ecchymosis feature extraction and recognition and tongue image acquisition device improvement.Proposed a tongue image segment algorithm based on Fully Convolutional Network(FCN),judged the position of the tongue based on gray projection,constructed VGG16-FCN-8s network for tongue image segmentation and used mathematical morphology to optimize the segmentation results.Proposed tongue ecchymosis extraction algorithm based on Otsu and watershed,extracted tongue ecchymosis by Otsu rough segmentation,watershed fine segmentation and postprocessing successively for subsequent analysis.Extracted location,shape and color features of ecchymoses,recognized shape and degree of ecchymoses by the extracted features and LIBSVM toolbox,proposed comprehensive ecchymosis index to quantify blood stasis of tongue different partions.Improved cameras,light source and intalled structure of tongue image acquisition device and developed camera control software based on MFC and EDSDK to set cameras parameters and control cameras.The experimental results show that the result of the segjmentation based on FCN was satisfactory.The ecchymosis extraction algorithm was effective and meet the analysis requirements.And the shape and degrer of ecchymosis were well reconized by the location,shape and color features.The proposed comprehensive ecchymosis index can reflect blood stasis on tongue.The improved tongue image acquisition device realized high quality tongue images acquisition.